RESUMO
Respondent-driven sampling (RDS) is a popular method for surveying hidden populations based on friendships and existing social network connections. In such a survey the underlying hidden network remains largely unknown. However, it is useful to estimate its size as well as the relative proportions of surveyed features. The fact that linked network participants are likely to share common features is called homophily, and is an important property in understanding the topology of social networks. In this paper we present a methodology that scales up RDS data to model the underlying hidden population in a way that preserves multiple homophilies among different features. We test our model using 46 features of the population sampled by the SATHCAP RDS survey. Our network generation methodology successfully preserves the homophilic associations in a randomly generated Barabasi-Albert network. Having created a realistic model of the expanded SATHCAP network, we test our model by simulating RDS surveys over it, and comparing the resulting sub-networks with SATHCAP. In our generated network, we preserve 85% of homophilies to under 2% error. In our simulated RDS surveys we preserve 85% of homophilies to under 15% error.
Assuntos
Inquéritos e Questionários , HumanosRESUMO
Performing network-based analysis on medical and biological data makes a wide variety of machine learning tools available. Clustering, which can be used for classification, presents opportunities for identifying hard-to-reach groups for the development of customized health interventions. Due to a desire to convert abundant DNA gene co-expression data into networks, many graph inference methods have been developed. Likewise there are many clustering and classification tools. This paper presents a comparison of techniques for graph inference and clustering, using different numbers of features, in order to select the best tuple of graph inference method, clustering method, and number of features according to a particular phenotype. An extensive machine learning based analysis of the REGARDS dataset is conducted, evaluating the CoNet and K-Nearest Neighbors (KNN) network inference methods, along with the Louvain, Leiden and NBR-Clust clustering techniques. Results from analysis involving five internal cluster evaluation indices show the traditional KNN inference method and NBR-Clust and Louvain clustering produce the most promising clusters with medical phenotype data. It is also shown that visualization can aid in interpreting the clusters, and that the clusters produced can identify meaningful groups indicating customized interventions.